• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# ANGLE Development Update - July 4, 2012
2
3We haven't posted an update on the development status of ANGLE in quite some
4time and we'd like to provide an update on some of the new features and
5improvements that we've been working on.
6
7## Conformance
8
9As announced in the [Chromium Blog]
10(http://blog.chromium.org/2011/11/opengl-es-20-certification-for-angle.html),
11ANGLE v1.0 has passed the Khronos OpenGL ES 2.0 certification process and is now
12a [conformant](http://www.khronos.org/conformance/adopters/conformant-products/)
13OpenGL ES 2.0 implementation.
14
15## Extensions
16
17We have recently completed the implementation of depth texture support
18([ANGLE\_depth\_texture]
19(https://code.google.com/p/angleproject/source/browse/extensions/ANGLE_depth_texture.txt?name=master))
20and earlier in the year we added support for instancing via attribute array
21divisors ([ANGLE\_instanced\_arrays]
22(https://code.google.com/p/angleproject/source/browse/extensions/ANGLE_instanced_arrays.txt?name=master)).
23See ExtensionSupport for a complete list of extensions that are supported by
24ANGLE.
25
26## Shader Compiler
27
28We have also made a number of improvements in the shader compiler.
29
30* We addressed a number of defects related to scoping differences between HLSL and
31GLSL and improved the scoping support in ANGLE's compiler front-end. We also
32worked with The Khronos Group to get an ESSL spec bug fixed and several items
33clarified.
34* We addressed a number of correctness issues in the GLSL to HLSL
35translation process. We fixed some bugs related to constant propagation and
36comma conditional assignments. More importantly, we fully implemented support
37for short-circuiting boolean logic operations. In GLSL, Boolean expressions do
38short-circuit evaluation as in C, but HLSL evaluates them entirely. This only
39has an observable effect if a short-circuited operation has side effects, such
40as a function call that modifies global variables.
41* We implemented detection
42for discontinuous gradient or derivative computations inside loops and replace
43them with explicitly defined continuous behaviour. HLSL and GLSL differ in their
44specified behaviour for operations which compute gradients or derivatives.
45Gradients are computed by texture sampling functions which don't specify a
46specific mipmap LOD level, and by the OES\_standard\_derivatives built-in
47functions. To determine the gradient, the corresponding values in neighbouring
48pixels are differentiated. If neighbouring pixels execute different paths
49through the shader this can cause a discontinuity in the gradient. GLSL
50specifies that in these cases the gradient is undefined. HLSL tries to avoid the
51discontinuity in the compiler by unrolling loops so that every pixel executes
52all iterations. This can make the D3D HLSL compiler spend a long time generating
53code permutations, and possibly even fail compilation due to running out of
54instruction slots or registers. Because the GLSL specification allows undefined
55behaviour, we can define such texture sampling functions to use mipmap LOD level
560, and have the derivatives functions return 0.0. To do this we examine the GLSL
57code's abstract syntax tree and detect whether the shader contains any loops
58with discontinuities and gradient operations. Within such loops, we generate
59HLSL code that uses explicitly defined texture LODs and derivative information.
60One additional consideration is that within these loops there can be calls to
61user-defined functions which may contain gradient operations. In this case, we
62generate variants of user-defined functions where these operations are
63explicitly defined. We use these new functions instead of the original ones in
64loops with discontinuities.
65
66These fixes result in ANGLE being able successfully compile a number of the more
67complex shaders. Unfortunately there are still some complex shaders which we
68have not yet been able to obtain solutions for. Ultimately Direct3D 9 SM3
69shaders are more restricted than what can be expressed in GLSL.  Most of the
70problematic shaders we've encountered will also not compile successfully on
71current ES 2.0 implementations.  We would only be able to achieve parity with
72Desktop GL implementations by using Direct3D 10 or above.
73
74## Texture Origin Changes
75
76We have also made a major change to ANGLE in the way the origin difference
77between D3D and OpenGL is handled. This difference is normally observable when
78using render-to-texture techniques, and if not accounted for, it would appear
79that images rendered to textures are upside down. In recent versions of ANGLE
80(r536 (on Google Code)-r1161 (on Google Code)), we have been storing surfaces
81following the D3D Y convention where (0, 0) is the top-left, rather than GL's
82bottom-left convention. This was done by vertically flipping textures on load
83and then adjusting the texture coordinates in the shaders to compensate. This
84approach worked well, but it did leave the orientation of pbuffers inverted when
85compared to native GL implementations. As of ANGLE r1162 (on Google Code), we
86have changed this back to the original way it was implemented - textures are
87loaded and stored in the GL orientation, and the final rendered scene is flipped
88when it is displayed to a window by eglSwapBuffers. This should be essentially
89transparent to applications except that orientation of pbuffers will change.  In
90addition to fixing the pbuffer orientation, this change:
91
92* eliminates
93dependent-texture look-ups in the shaders, caused by flipping the texture
94y-coordinates
95* rounding of texture coordinates (while previously within spec)
96will be more consistent with other implementations, and
97* allows potential
98faster paths for loading texture data to be implemented. The only potential
99downside to this approach is that window-based rendering may be a bit slower for
100simple scenes. The good news is that this path is not used by browser
101implementations on most versions of Windows.
102
103## Preprocessor
104
105Finally, Alok P. from Google has been working on implementing a new shader
106preprocessor for the last number of months and this effort is nearly complete.
107This new preprocessor should be more robust and much more maintainable. It also
108includes many (~5000) unit tests and passes all WebGL conformance tests. If you
109wish to try this out before it is enabled by default, define
110ANGLE\_USE\_NEW\_PREPROCESSOR=1 in your project settings for the
111translator\_common project.
112
113## Contributions
114
115As always we welcome contributions either in the bug reports (preferably with an
116isolated test-case) or in the form of code contributions. We have added a
117[ContributingCode](ContributingCode.md) wiki page documenting the preferred
118process for contributing code. We do need to ask that you sign a Contributor
119License Agreement before we can integrate your patches.
120